Ancestry-based differences in the immune system are associated with lupus severity

Systemic lupus erythematosus (SLE) affects 1 in 537 of African American (AA) women, which is >2-fold more than European American (EA) women. AA patients also develop the disease at a younger age, have more severe symptoms, and a greater chance of early mortality. We used a multi-omics approach to uncover ancestry-specic immune alterations in SLE patients and healthy controls that may contribute to disease disparities. Cell composition, signaling, and epigenetics were evaluated by mass cytometry; droplet-based single cell transcriptomics and paired proteogenomics (scRNA-Seq/scCITE-Seq). Soluble mediator levels were measured in plasma and stimulated whole blood. Toll-like receptor (TLRs) pathways are activated by vaccination and microbial infection, and are also key drivers of autoimmune disease. We observed enhanced TLR3/4/7/8/9-related gene expression in immune cells from AA versus EA SLE patients. TLR7/8/9 and IFN α phospho-signaling responses were heightened even in immune cells from healthy AA versus EA controls. TLR stimulation of healthy AA and EA immune cells recapitulated the distinct ancestry-associated SLE immunophenotypes. Thus, healthy individuals show ancestry-based differences in innate immune pathways that could in�uence the course and severity of lupus and other diseases.


Introduction
Systemic lupus erythematosus (SLE) is a heterogenous disease with varied environmental and genetic risk factors that may contribute to risks for development of the disease and to its disease course (1,2). SLE prevalence is 2-3 times higher in African Americans (AA) than European Americans (EA) (3). AA ancestry is associated with greater disease activity, severity, and worse outcomes in SLE with a higher prevalence of end-stage renal failure and earlier mortality (2). Lupus remains a top 10 medical cause of death in AA women between 15 and 45 years of age (4), and individuals of African ancestry have a more abrupt and earlier age of SLE onset compared to those of EA descent (5).
Toll-like receptors (TLRs) 7 and 9 are pattern recognition receptors that sense single-stranded RNA and DNA, respectively, and are critical for type I IFN production and the generation of anti-viral immune responses. Activation of TLR7 and TLR9 by recognition of self nucleic acids is implicated in the pathogenesis and development of autoimmune disease (6). In lupus mouse models, TLR7 is a critical driver of disease, including nephritis, and anti-RNA antibodies (7), and TLR9 is important for the development of anti-DNA antibodies (7)(8)(9). Human lupus is associated with increased expression and activation of TLR7 and TLR9 on peripheral blood mononuclear cells, and activation of these TLRs induces IFNα production (10)(11)(12)(13). Certain pathways driving lupus pathogenesis are known to be enhanced in AA versus EA populations, with greater frequencies of patients with elevated IFNα, B cell activation markers (CD86), and double negative 2 (DN2) autoantibody producing B cells (14)(15)(16). Patient ancestry has been shown to in uence gene expression signatures which help de ne the molecular heterogeneity of SLE (17), with B cell-related expression signatures more heavily perturbed in AA versus EA, and greater plasma cell signatures found in AA SLE patients (17). Differences in immune cell frequencies of healthy AA versus EA individuals have also been described, such as benign neutropenia, a greater CD8/CD4 ratio, and higher lymphocyte numbers, primarily B cells (18,19). B cell numbers also differ by sex, with the highest cells/mL reported in AA females (20).
Additionally, macrophages from AA individuals display a stronger in ammatory response to pathogens such as Listeria, Salmonella, and In uenza when compared to macrophages from EA individuals (21,22).
Given ancestry-related differences in the immune response of healthy individuals, the current study investigated ancestry in uences on immune response pathways linked to lupus, including TLR7, TLR9, and IFNα using a multi-omic, systems level approach. Ancestry-associated differences in dominant TLR signaling pathways that may in uence lupus disease phenotype and severity were found in both healthy controls and SLE patients, suggesting the likelihood of inherited autoimmune-permissive immunologic patterns.

Cohorts of AA and EA populations for immune analysis
To examine relationships between immune phenotypes, ancestry, and disease activity, the following were evaluated: 1) healthy controls (n=10 EA, n=8 AA), 2) SLE patients with inactive disease (SLE INACT: SLEDAI<4, n=10 EA, n=10 AA), and 3) SLE patients with active disease (SLE ACT: SLEDAI≥4) (n=9 EA, n=11 AA) ( Fig. 1 and Fig. S1). Patients were matched by self-reported ancestry, SLEDAI, sex, and medication use (Table S1). AA SLE ACT patients tended to be younger in age and there were no observable differences by ancestry in SLE INACT patients.
Clinical criteria were similar across SLE INACT patients of both ancestries. SLE-ACT patients displayed ancestry associated differences in the proportion with DNA binding autoantibodies (AA 60% vs EA 33%), leukopenia (AA 30% and EA 0%), and lymphopenia (AA 50% vs EA 33%) (Table S2). Autoantibody speci cities differed slightly between AA and EA patients with AA patients having higher frequencies of RNP (26.3% EA vs 52.4% AA) or Sm/RNP (31.6% EA vs 61.9% AA) speci c antibodies, and EA patients having higher frequencies of Ro (52.6% EA vs 23.8% AA) and La (15.8% EA vs 4.8% AA) speci c antibodies. The number of speci cities was higher in SLE ACT versus SLE INACT patients of both ancestries (Table S3-

Lupus immune phenotypes vary by ancestry and disease activity
Mass cytometry was performed to investigate whole blood phenotypes and phospho-signaling (Fig. 1A).
To assess changes in TLR cell signaling pathways, the mass cytometry panel was designed with 33 metal isotype-tagged monoclonal antibodies speci c for cell lineage, phospho-proteins, activation, and homing markers that discern major immune cell subsets and subpopulations (Table S5). Single cell populations were visualized using a high-dimensionality reduction method (tSNE) that down-sampled the total cell population and unbiasedly clustered cells by likeness. Frequencies were determined by a standard biaxial-gating scheme that incorporated all intact, live cells collected (Fig. 2A). The tSNE analysis included over 1.3 million cells (~17,000 cells/sample) and distinguished 22 phenotypically distinct clusters (Fig S3A). Marker expression of gated phenotype clusters is summarized for major lineage markers via tSNE plot (Fig. S3B) and by median intensity in a heatmap (Fig. S3C).
Concatenated les of all individuals in each disease group were used to create tSNE images, demonstrating that the immune cell composition of whole blood was different in all six cohorts, depending on both ancestry and disease scores (Fig. S3D). For instance, B cell and non-classical monocyte frequencies were greater in AA versus EA controls ( Fig. 2A, Table S6-7). T cells, primarily CD8+ T cells, were also higher in SLE INACT AA versus EA patients. In contrast, neutrophil counts trended lower in AA versus EA controls, as expected ( Fig. 2A) (18)(19)(20).
Increased disease activity in EA patients was associated with reduced frequencies of B cells, naïve and double negative (DN) (CD27-IgD-) B cells, CD4+ T cells, DN T cells, and basophils/eosinophils ( Fig. 2A, Table S6-7). In AA samples, more severe disease was associated with reduced frequencies in nonclassical monocytes, NK cells, CD8+ T cells, and CD141+ dendritic cells frequencies, as well as increased frequencies of memory B cells/plasmablasts ( Fig. 2A, Table S6-7, Fig. S4). These data demonstrate that the immune compositions of whole blood from AA and EA populations differ in meaningful ways during SLE disease progression, and re ect some differences already present in healthy people.
Cell activation is higher in AA versus EA SLE patients We next analyzed activated and in ammatory immune populations in the six cohorts. Brie y, we assessed the frequencies and mean metal intensity (MMI) in whole blood cell subsets by monitoring CD38, HLA-DR, CD11c, CD11b, and CD27 (Table S8- 9). We found that CD38 antigen (Fig. 2B) and CD38+HLA-DR+ co-expressing activated T cells, monocytes, pDCs, and NK cells were elevated with increased disease activity in all SLE patients ( Fig. S5A-H).
The T cell co-stimulatory molecule CD27 assists in antigen-speci c expansion and supports survival of activated T cells (23). CD27, was more highly expressed in CD8+ T cells ( Fig. 2C-D) and CD4+ T cells ( Fig. 5I), across all AA versus EA samples. In addition, CD38 was more highly expressed in CD8+ T cells, DN T cells, and pDCs in AA versus EA SLE ACT patients ( Fig. 2E-F, Fig. S5J).
Using markers in this panel, we did not observe changes in the frequencies of activated B cells between SLE ACT and SLE INACT disease activity (Table S8-9). However, AA versus EA naïve B cells expressed higher levels of IgD, even in the absence of disease (Fig. S5K-L). The immunoglobulin IgD is elevated in rheumatic disease, has a high proclivity toward binding to nuclear antigens, and induces potent in ammatory responses in myeloid cells (24,25). In addition, the frequency of activated neutrophils increased with disease activity in AA patients, but not in EA patients (Fig. 2G-H).
To examine associations between cell populations and disease activity, linear regression analyses were performed with SLEDAI versus cell subset frequencies. In EA SLE patients, naïve B cell frequencies were negatively associated with disease activity (R 2 =0.2014) (Fig. S6A). In AA SLE patients, T cell frequencies (both CD8+ T cells (R 2 =0.2286)) and CD4+ T cells (R 2 =0.2094)) were negatively associated with disease activity, whereas activated CD38+ T cells were positively associated ( Fig. S6B-D). In summary, CD8+ and DN T cell activation, pDC, and neutrophil activation were more prevalent in AA SLE ACT patients compared to EA SLE ACT patients. We infer that the activated and in ammatory immune populations in whole blood differ between AA and EA populations, both in healthy individuals and during SLE disease course.
H3K27ac in CD8+ T cells is higher with AA ancestry and disease severity To investigate a potential relationship between immune cell epigenomes, ethnicity, and disease activity, we performed epigenetic landscape pro ling using cytometry by time-of-ight (EpiTOF) (26,27). The global levels of 40 chromatin modi cations were measured on a single-cell level in 19 different cell populations (Fig. S7). Individuals were split into two biological replicates for testing (Fig. S8).
In most immune cell populations, diverse chromatin modi cations were elevated in SLE patients compared to controls (Fig. 3A). However, pDCs, effector CD4+ and effector CD8+ T cells from SLE patients had lower levels of most chromatin modi cations. SLE patients could be distinguished from healthy controls with high sensitivity (>85%) and speci city (>70%) using just three chromatin marks (H3K4me2, H3K4me3, and H4K20me1) (Fig. S9). These three chromatin marks had equal discriminatory capacity in both biological replicates.
We compared chromatin marks on 19 different immune cell populations between SLE patients and healthy controls using both AA and EA samples (Fig. 3B). The levels of H3K18ac (Fig. 3C) and H3K27ac ( Fig. 3D) were higher in CD8+ T cells on all AA versus EA samples. Samples from AA patients could be distinguished from EA samples with high sensitivity and speci city (AUC=80%) using H3K27ac levels in CD8+ T cells alone (Fig. 3E). Further, H3K27ac in CD8+ T cells was higher in AA SLE ACT patients compared to SLE INACT patients, but this was not observed in EA patients (Fig. 3F). This was the only marker found to signi cantly distinguish disease activity (data not shown). These ndings suggest that immune cell epigenomes of EA and AA populations are different, both in healthy individuals and during different SLE disease activity states.

IgG+ B cells are more frequent in AA SLE ACT patients
To investigate the transcriptional pro les of antigen-presenting cells in AA and EA SLE patients, we performed droplet-based scRNA-sequencing (10X Genomics) and protein genomics (CITE-seq) using CD2 depleted PBMCs on all study participants (Fig. 1C). After ltering out cells of low quality and batch normalization ( Fig. S10-11), we obtained transcriptome datasets from ~90,000 cells, with over 30 million unique transcripts from PBMCs of all samples. Among these cells, 31,082 cells (35.3%) were from controls, 29,838 cells (33.9%) were from SLE INACT patients, and 27,133 cells (30.8%) were from SLE ACT patients (Fig. S12). Using graph-based clustering of uniform manifold approximation and projection (UMAP), we captured the transcriptomes of 8 major cell types or subtypes according to the expression of canonical gene and protein markers (Fig. S13-14).
We used protein and transcriptome data to subdivide B cells into six subsets based on lineage markers ( Fig. 4A-C). For both ancestries, increased disease activity was associated with a reduced frequency of naïve B cells (Fig. 4D). An increased frequency of age-associated B cells (ABCs), germinal center (GC) B cells, and plasmablasts ( Fig. 4E-G), but not transitional or memory B cells (Fig. 4H-I), was also found. The reduced frequencies of naïve B cells in SLE ACT patients was also observed by mass cytometry, and coincided with increased gene expression of B cell differentiation, proliferation, and migration pathways, as well as decreased expression in apoptotic pathways (Fig. S15, Fig. 4J-K). IFN and RNA viral infection gene expression pathways were elevated in plasmablasts from patients with high disease activity ( Fig.  S16-17). Increased class-switching was also evident in lupus patients with higher disease activity ( Differences in IFN gene expression signatures in classical, non-classical, and intermediate cell subsets were assessed using prede ned gene expression modules (28, 29). EA patients expressed similar elevated levels of IFN gene expression modules regardless of disease activity (Fig. 5H). In contrast, IFN pathway involvement in monocytes increased in AA SLE patients with active disease, particularly in intermediate monocytes (Fig. 5H). Apoptosis pathway involvement was reduced in intermediate monocytes from SLE ACT patients (Fig. S19H). These data are consistent with the hypothesis that expanded ISG Hi monocyte populations contribute to heightened disease activity in both EA and AA patients, while greater IFN activation in intermediate monocytes in AA SLE ACT patients may accentuate activation and increased migration into affected tissues.
Transcriptionally de ned dendritic cells could be split into three distinct subsets of CLEC9a+ conventional DCs (cDCs), CD1c+ cDCs, and pDCs ( Fig. 5I-K, Fig. S19I-J). CD1c+ cDCs were elevated with disease severity in EA subsets, whereas pDCs were higher in all AA versus EA subsets. HLA Class I and II antigen presentation transcripts were differentially expressed in cDCs of SLE patients (Fig. 5L). AA SLE ACT patients had the highest levels of HLA Class I expression compared to all other groups (Fig. 5L), whereas, EA SLE patients trended higher in expression levels of multiple HLA Class II transcripts. These data reveal transcriptome differences between dendritic cells of AA and EA individuals, both in health and during SLE disease progression. TLR7/9 and IFNα signaling pathways are elevated in AA immune cells Given that Toll-like receptors (TLRs) 7 and 9 and Type I Interferon are key drivers of murine autoimmune disease, we investigated whether differences in TLR and IFN signaling pathways underlie the observed ethnic differences in immune phenotypes, in health and SLE. To assess differences in TLR activity, Ingenuity Pathways Analysis (IPA) was rst used to evaluate differentially expressed genes in monocyte and B cell clusters identi ed by UMAP of scRNA-sequencing ( Fig. 6A-B, Table S10). TLR gene expression pathways were elevated in naïve B cells, memory B cell, and ABCs from SLE ACT AA patients versus EA patients (Fig. 6A). TLR gene expression pathways were also elevated in monocytes from AA versus EA SLE ACT patients, most notably among more activated/in ammatory monocyte populations: intermediate, CXCL8 Hi , polyfunctional, and CD14 Low CD16 low monocytes (Fig. 6B). The TLR4 signaling pathway (LPS) showed the greatest elevation, as previously reported (21), but TLR3, TLR7/8, and TLR9 pathways were all elevated in AA versus EA SLE ACT patients.
Next, 9 signaling proteins and 8 activation markers were monitored to assess TLR signaling. Phospho-CyTOF was performed before and after a 4 or 15 minute stimulation of whole blood (Fig. 6C, Table S11-18, Fig. S20-28), using cell surface markers to discern 8 major cell lineages (Fig. S20). IFNα mediated induction of pSTAT5 was diminished in most cell lineages from AA patients versus AA healthy controls ( Fig. S29-30). TLR7/8 and TLR9 induced phospho-signaling markers and activation proteins were also lower in multiple cell lineages from AA SLE patients compared to AA healthy controls ( Fig. S29-33). Lower fold changes were not a result of SLE patients cells being maximally stimulated, as they had lower total levels of phospho-signaling proteins following stimulation (Fig. S30B). TLR7/8-and TLR9-mediated induction of p-p38 and pERK1/2 were also lower in EA patients versus healthy controls, which may suggest exhaustion of these pathways from persistent stimulation in patients.
There were signi cant differences in TLR signaling responses between AA and EA healthy controls. For instance, relative to EA, AA samples showed increased IFNα-mediated induction of pSTAT5 as well as increased PMA/Ionomycin-, TLR7/8-and TLR9-mediated induction of cCASP3, pCREB, pERK1/2, pPLCy2, Syk, and p-STATs (Fig. 6C). The TLR9 responses of B cells from healthy AA versus healthy EA were particularly more prone to activation, with higher expression of CD38, CD27, HLA-DR, CD11c and CD11b after stimulation. These data suggest that TLR7/8/9 and IFNα signaling pathways are inherently different in healthy AA versus EA individuals, which may in uence lupus disease manifestations and severity.
TLR stimulation of healthy AA and EA immune cells recapitulates ancestry-associated SLE immunophenotypes AA SLE ACT versus SLE INACT patients have fewer NK cells, T cells, non-classical monocytes, and more memory B cells/plasmablasts ( Fig. 2A), whereas EA SLE ACT versus SLE INACT patients have fewer B cells ( Fig. 2A). We hypothesized that the observed ancestry-dependent differences in TLR7, TLR9, and IFN signaling in healthy controls might contribute to the differences in the immune cell frequencies of SLE patients.
To test this hypothesis, PBMCs were obtained from 10 AA and 10 EA healthy controls (matched by age, sex, and BMI), and cells were stimulated in vitro with TLR7/8 agonist, TLR9 agonist, or IFNα alone or in combination, for 7 days. All controls were female, had a mean age of 29.2 years, and no clinical symptoms of autoimmune disease as determined by a connective tissue screening questionnaire (CSQ) (30), were anti-nuclear antibody negative (ANA) by Hep2 immuno uorescence and by Bioplex 2200, and had no anti-cyclic citrullinated peptide (CCP) or rheumatoid factor (RF)-speci c autoantibodies (data not shown).
The immune cell compositions following stimulation of PBMCs with TLR7/8 (R848) and/or TLR9 (CpG) agonists were signi cantly different for AA versus EA healthy controls (Fig. 7A-B). In particular, the combination of TLR7 and TLR9 agonists resulted in increased T cells (Fig. 7C) and NK cells (Fig. 7D) in AA versus EA healthy controls, but no differences in total B cells (Fig. 7E) and myeloid cells (Fig. 7F). B cells were then classi ed into 6 subsets using CD27 and IgM gating via biaxial plots (Fig. 7G). Simultaneous treatment with TLR7 and TLR9 agonists resulted in fewer naïve B cells and increased class-switched plasmablasts in AA versus EA healthy controls (Fig. 7H-M). AA samples produced more DN B cells in response to R848, R848+IFNα, and R848+CpG+IFNα (Fig. 7I), and more class-switched memory B cells and class-switched plasmablasts in response to R848+CpG (Fig. 7J-K). IgG levels were also higher in AA versus EA, but no differences were observed in IgA or IgM following TLR7/8, TLR9, or IFNα stimulation ( Fig. 7L-M, Fig. S34). The response to simultaneous TLR7/8 and TLR9 stimulation in AA samples induced a stronger T cell, NK cell, and class-switched B cell response, whereas EA individuals had predominantly more naïve B cells. These changes recapitulate the cell population changes seen in AA and EA SLE patients.
Elevated pro-in ammatory cytokines in AA SLE patients and altered responses to TLR-driven cytokine pathways in lupus patients Next, we tested whether increased disease activity and altered immune responses in EA and AA SLE patients were associated with alterations in chemokine and soluble mediator pro les. We assessed 39 different mediators by multiplex bead-based assays or ELISA (Fig. 1B) and found that plasma from AA SLE patients had increased levels of pro-in ammatory cytokines and other in ammatory regulators when compared to plasma from EA SLE patients ( Fig. 8A-G, Fig. S35A-G, Table S19-20). The chemokine, Eotaxin, which helps to recruit eosinophils, was the only cytokine that was signi cantly increased in EA versus AA patients (Fig. 8H). Stem cell factor (SCF or c-kit) and MCP-1 levels best correlated with disease activity (Fig. 8I-J). These data suggest that higher levels of pro-in ammatory cytokines contribute to altered and enhanced disease activity in AA SLE patients.
To determine whether dysregulated signaling pathways contribute to the differences in soluble mediators, we stimulated whole blood from controls, SLE INACT and SLE ACT patients with vehicle, R848/CpG/LPS (TLR), PHA, or PMA+Ionomycin for 24 hours, collected supernatants and measured the levels of 39 soluble mediators (Fig. 1A-B, Fig. S36). Cytokine levels were already signi cantly elevated in unstimulated culture supernatants of SLE patients compared to controls, most notably in AA SLE ACT patients ( Fig.  S35H-L). Following TLR stimulation of SLE ACT whole blood, a subset of pro-in ammatory cytokines was increased, including CXCL13, IL-13, IL-6, SCF, and TNF-β, with trends higher in most other cytokines (Fig. 8K, Fig. S37, Table S21-22). In contrast, TLR-mediated stimulation of Type I and Type II IFN cytokine pathways were speci cally reduced in SLE patients versus healthy controls, particularly for SLE ACT patients ( Fig. 8L-N). Diminished Type I IFN responsiveness was speci c to TLR stimulation, whereas Type II IFN reduction were also observed following PHA and PMA-Ionomycin stimulation ( Fig. S38-40, Table   S23-26). The reduction in TLR-induced IFNα levels was associated with higher IFN gene signatures in classical and non-classical monocytes, suggesting that this may be due to exhaustion of the TLR-IFN cytokine pathway ( Fig. 8O-P). The diminished IFN response was not signi cantly different between EA or AA patients; however, EA SLE ACT patients produced higher levels of the regulatory cytokines IL-1RA and IL-10, which may mitigate TLR-induced in ammation ( Fig. S37B-C). The altered TLR response in lupus patients, with diminished early IFN production and increased pro-in ammatory cytokine levels, indicates external or internal stressors that trigger the TLR pathway may be associated with altered responses to infection or induction of disease are. These alterations are more severe in patients with increased disease activity.

Discussion
In this study, we identify a signi cant impact of ancestry on immune composition and cellular signaling responses triggered by viral pathways. The likelihood that this shapes the immune phenotype in SLE is supported by the markedly stronger response of AA healthy controls to TLR7/8, TLR9, and IFNα stimulation pathways known to drive lupus pathogenesis. These results were consistent across multiple platforms with higher TLR gene expression pathways, greater cell activation and phospho-signaling pro les, and ampli ed in vitro responses seen in AA individuals. Past studies have demonstrated increased pro-in ammatory responses to bacterial pathogens and in uenza in AA patients (21,22), higher rates of in ammatory diseases (31), and greater lupus and COVID-19 disease severity (2,32,33). We speculate that stronger TLR7/8/9 responses contribute to greater disease burden, earlier age of onset, and worsened disease symptoms known to occur in AA SLE patients.
Ancestry-related differences in cell composition and the TLR7/8/9 and IFNα responses may be explained by a combination of genetic and environmental factors (22). Benign neutropenia, which helps account for lower frequencies of neutrophils and higher lymphocyte counts in the circulation of AA individuals, has been linked to rs2814778 (G) variant of the gene encoding atypical ACKR1, the duffy antigen receptor for cytokines (DARC) (34). It is possible that higher frequencies of TLR7 and TLR9 expressing cells in the circulation may contribute to greater expansion of these populations in response to stimuli in AA. Higher expression of TLR7 and TLR9 have been reported on AA versus EA SLE patient PBMCs (13). Women of AA descent were also found to carry signi cantly more allelic variants known to upregulate proin ammatory cytokines that may in uence the strength of TLR7/8/9 responses (35). Environmental factors, such as exposure to Epstein-Barr virus and cytomegalovirus trigger the same TLR pathways, and may further prime immune responses to favor increased pro-in ammatory cytokine production. AA individuals also have a disproportionally higher prevalence of EBV, CMV, and HSV-1 antibodies that associate with socioeconomic variables (36).
One of the differences in the immune response of AA SLE ACT patients was increased plasmablast, memory B cell, ABC, and germinal center B cell involvement contributing more evidence to a stronger B cell phenotype in AA SLE patients (14)(15)(16)(17)(37)(38)(39). However, we found that this phenotype is not unique to the SLE patients in the AA population. Healthy AA controls generated more IgG class-switched memory B cells and plasma cells in response to TLR7/8 and TLR9 stimuli. This expansion of IgG+ B cells, and the ability to class-switch to IgG, requires at least 3 cell divisions, the presence of T cell stimulation, and a proper cytokine milieu that includes IL-13, IL-4, and/or IL-10 (40). We found IL-13 and IL-4 in higher concentrations in the plasma of AA SLE patients with greater T cell activation and CD38 and CD27 costimulatory marker expression in AA SLE patients.
In addition to an altered B cell axis in AA versus EA SLE patients, a more extensive AA-associated SLE phenotype was found that included greater CD8+ T cell, NK cell, non-classical monocyte, and DC involvement. We previously found T cells to be increased in SLE patients, with greater numbers in AA SLE patients, and increased T cell activation pro les in AA patients prior to lupus development (41).
In this study, simultaneous treatment with TLR7/8 and TLR9 agonists drove increased T cells in AA healthy control cultures. Effective activation of naïve T cells requires three distinct signals: recognition of peptide-HLA complex on APCs, co-stimulatory molecule interaction of the T cell and APC, and proin ammatory cytokine signals from the external environment. Dendritic cells in AA SLE ACT patients express more β-2 microglobulin (B2M) and other HLA class I molecules, including HLA-A, HLA-B, and HLA-C, compared to EA SLE ACT patients and AA controls and SLE INACT patients that may increase HLA class I self-peptide presentation. The T cells of AA samples exhibited higher expression of CD27, a costimulatory marker that supports antigen-speci c expansion and sustains activation. Further, elevated levels of cytokines TNFα, IL-4, IL-13, SCF and others in AA patients can increase T cell differentiation and expansion. A skewing of the T cell response with decreases in CD8+ T cells in AA SLE ACT versus CD4+ T cells in EA patients was also observed. Decreases in effector cell populations with disease onset have been previously seen in autoimmune disease as activated cells migrate to affected tissues (42)(43)(44).
Heightened gene expression of transmigratory transcripts, increased activation markers, and decreased apoptosis transcripts in these cell populations suggest cells become activated and leave the periphery.
T cells were correlated with SLE disease activity in AA patients. CD8+ T cell epigenetic modi cations in H3K18ac and H3K27ac, both associated with higher activation of transcription, was enhanced in AA SLE ACT patients. Our previous work has shown, that H3K27ac is correlated with cytokine production (45), namely TNFα, which this study found was also elevated in AA versus EA SLE patients. We found that treatment with the histone acetyl transferase inhibitor A-485 led to a concentration-dependent decrease in global histone H3K27ac levels that did not affect cell viability (45). More importantly, treatment with A-485 led to a major diminution in the frequency of IL-1β and TNFα suggesting this may be an effective target in AA SLE patients.

CD8+ T cells are also the predominant in ltrating immune cell subset In SLE patients with class III or IV
nephritis, which is more prevalent in AA patients (46). Further, single cell transcriptional analyses revealed NK cells and ABCs to signi cantly contribute to immune in ltrates in the kidney (47). Accumulation of CD8+ T cells in periglomerular areas was associated with worse outcome and greater disease severity (46, 48). Heightened expression of HLA Class I in DCs and increased frequencies of in ammatory nonclassical/intermediate monocytes in AA versus EA SLE ACT patients may direct the immune response towards more auto-aggressive CD8+ T cells due to increased cross-presentation (49,50). Higher SCF levels in the plasma may also contribute to increased dendritic cell activation and T cell differentiation (51). Stronger TLR7/8/9 and IFNα signaling that drive heightened CD8+ T cell, NK cell, monocyte/DC, and B cell responses likely contribute to a higher risk for lupus nephritis in AA patients.
While heightened TLR gene expression pathways were found in SLE patients, phospho-signaling responses following whole blood TLR7/8 and TLR9 stimulation were reduced. This is consistent with previous reports that correlated diminished responses to transcriptional signatures of cellular senescence, stress, and shortened telomere lengths (52). In the current study this evidence of exhaustion is enhanced in AA versus EA patients with heightened disease activity. Hastened telomere shortening is known to contribute to earlier aging and mortality in SLE patients (53). Stronger TLR7/8/9 signaling pathways that drive exhausted responses, stress, and telomere shortening may lead to earlier mortality in AA SLE patients.
Lupus patients, speci cally those with heightened disease activity and lupus nephritis, are known to be particularly prone to infections, which are a leading cause of morbidity and mortality in SLE (54-56). While part of the high infection risk is likely due to immunosuppressive medications, alterations in the immune response due to lupus may also contribute to heightened risk. Following stimulation with TLR agonists, lupus patients with high disease activity produce signi cantly less Type I and II IFN and IFNassociated cytokines regardless of ancestry. This decreased IFN response is directly associated with higher IFN gene signature scores in monocytes at baseline. Although IFN levels were reduced compared to controls, other pro-in ammatory cytokine levels remained high, a phenotype consistent with aberrant and severe immune responses in certain viral infections (57, 58). EA SLE ACT patients were distinguished from AA SLE ACT patients by higher levels of the regulatory cytokines IL-10 and IL-1RA that may help dampen pro-in ammatory in ammation in the context of a dysregulated immune response. Exhausted IFN pathways from chronic in ammation in lupus patients may contribute to greater infection risk, more severe manifestations, and lead to the lupus ares that are more common in AA patients.
TLR7 and TLR9 pathway activation is also required for effective responses to vaccines and viral infections, such as COVID-19, which disproportionately affects AA and Hispanics more severely than EA.
Severe COVID-19 infection is characterized by a stronger CD8+ T cell response, whereas mild COVID-19 patients have reduced frequencies and a more exhausted CD8+ T cell phenotype (59). Further, severe COVID-19 patients have an increased and more highly activated NK cell response (60). Multi-system in ammatory syndrome in children (MIS-C), which occurs in some individuals after COVID-19 infection, is also more prevalent in AA and Hispanic children and is de ned by a lack of exhausted CD8+ T cells that maintain the in ammatory environment (61). AA individuals are reported to have higher antibody titers to rubella, measles, pertussis, and in uenza vaccinations (62-65). These phenotypes are consistent with the heightened response to TLR7/8 and TLR9 pathways observed in healthy AA controls, and are the same immune subsets altered during heightened SLE disease in AA patients. While these responses may be helpful in vaccination and promoting an effective memory response, it may predispose individuals of AA descent to more severe disease manifestations during infection and autoimmune disease.
Limitations of the current study include the small size and the cross-sectional study design, which allows capture of the immune pro le at only one timepoint. A secondary cohort with a larger sample size to con rm the ndings of this study including a longitudinal assessment of AA versus EA SLE patients with changes in disease activity over time will be important to test the robustness of these ndings. Despite the small sample size, the consistency of phenotype differences across multi-omic platforms helps substantiate the ancestry-based results that have come to light in this study, which may be critical clues in developing precision medicine for SLE. However, it is also important to acknowledge that ancestry alone is an imperfect way to approach individual phenotyping and therefore it is critical to avoid pigeonholing best medical practice in the future based only on racial background.
The results of this project provide new insight into the contribution of ancestry in driving immune composition and responses in SLE disease activity. Heightened TLR7/8, TLR9 and IFNα signaling pathways in the AA immune response support immune phenotypes seen in response to infection and vaccines, and in autoimmune disease. Increased CD8+ T cell, NK cell, B cell, monocyte, and DC activity in AA individuals after stimulation of TLR7/8/9 may contribute to increased disease severity and early mortality in AA SLE patients, and these pathways may be targeted for more personalized treatments. Our results stress the importance of ancestry assessment in the illumination of immune phenotypes that may be key to outcomes across medical elds.

Materials And Methods
Study Population and Sample Collection. All experiments using human samples were conducted in accordance with the Helsinki Declaration and approved by the Institutional Review Board at the Resquimod (R848)). A second aliquot of whole blood was diluted 1:2 (400 ul total) with complete RPMI (RPMI supplemented with 10% fetal calf serum, penicillin streptomycin, and glutamine) in cluster tubes.
Phospho-signaling Mass Cytometry Staining and Acquisition. Phospho-CyTOF was performed at the Human Immune Monitoring Center at Stanford University. Upon thawing, samples were washed and surface stained for 30 min at room temperature (Table S5). Cells were washed twice, permeabilized with 100% methanol and kept at -80C overnight. The next day, cells were washed and resuspended in intracellular antibody cocktail for 30 min before washing twice. Cells were resuspended in 100 mL iridium-containing DNA intercalator and incubated for 20 minutes. Cells were washed once with cyFACS buffer and twice with MilliQ water. Cells were diluted to 750x10 5 cells/mL in MilliQ water and acquired on CyTOF. Data analysis was performed using viSNE in Cytobank to determine cell populations, and fold change of the 95th percentile was calculated for all 9 signaling molecules and 8 major cell populations.
Epigenetic Landscape Pro ling Using Mass Cytometry (EpiTOF). As part of the ACE Network Collaborative Project, EpiTOF was performed as previously described (27) In Vitro TLR7, TLR9, and IFNα Stimulations with Healthy Controls. PBMCs from 20 healthy controls were recruited through health fairs, as previously described (41). 10 EA and 10 AA healthy individuals were selected for in vitro stimulation studies. To ensure no individuals had any autoimmune disease manifestations, only controls that were ANA negative by both IIF and Bioplex, negative for CCP and RF antibodies, and had scored 0 on CSQ in all autoimmune disease categories were selected. Individuals were all female and were matched across ancestries by age and BMI. Cells were plated at 1x10 6 cells/mL (900ul) in a 24 well plate, where 100 ul of stimuli was added in the following stimulation conditions: media (unstimulated), R848, CpG, IFNα, R848+IFNα, CpG+IFNα, R848+CpG, R848+CpG+IFNα. R848 and CpG were both used at a nal concentration of 300 ng/mL, while IFNα was used at 1000 U/mL. Cells were incubated with stimuli at 37 o C, 5% CO 2 for 7 days. Plates were spun down, supernatants were collected and stored at -80 o C until use for antibody ELISAs. Cells were collected for B cell and other cell subset assessment using an 8 marker antibody panel for ow cytometry. Cells were washed, incubated in                                  Table S15. pDCs-fold change in phospho-proteins following stimulation with various stimuli Table S16. NK cells-fold change in phospho-proteins following stimulation with various stimuli   Table S17. Monocytes-fold change in phospho-proteins following stimulation with various stimuli Table S18. Granulocytes-fold change in phospho-proteins following stimulation with various stimuli  Figure 1 Analysis of ancestry-associated differences in immune phenotypes. The work ow is broadly divided into 3 steps. First, 58 samples from EA and AA healthy controls, and SLE INACT and SLE ACT patients were matched by age, ancestry, and gender. (A) Whole blood was collected, left unstimulated or stimulated, and used for immunophenotyping by mass cytometry and signaling analysis by phospho-CyTOF (B) Plasma and supernatants collected after overnight stimulation of whole blood were collected for assessment of 39 different soluble mediators using multiplex bead-based assays and ELISAs. (C) PBMCs were used for droplet-based scRNA-seq and EpiTOF. For sequencing, PBMCs were washed and depleted of red blood cells and depleted of T cells using CD2 depletion to enrich for non-T cell populations. A 10X genomics platform was to examine differences in gene expression and a 51-plex CITE-seq panel was utilized to delineate speci c cell lineages, activation and regulatory markers. (D) PBMCs from EA and AA healthy controls with no autoimmune disease manifestations were stimulated for 7 days with IFNa, TLR7/8, or TLR9 agonists, alone or in combination, to assess immune composition and antibody production by ow cytometry and ELISA, respectively.     indicates increased TLR pathway activity in AA versus EA, white indicates no change, and blue indicates decreased TLR activity. (C) Peripheral whole blood from were stimulated for 4 or 15 minutes with IFNa, PMA and ionomycin, or TLR4, TLR7/8, or TLR9 agonists. The median 95 th percentile was used to calculate the fold change of phospho-signaling and activation markers in AA versus EA controls in 8 cell populations (see legend: B cells, CD4+ T cells, CD8+ T cells, dendritic cells (DCs), plasmacytoid DCs (pDCs), natural killer (NK) cells, monocytes (mono) and granulocytes (gran). Signi cant changes are noted by a blue box (decrease) or a red box (increase). p<0.05